The Money-Income Causality Debate

A Billion-Mark Note


"[Recessions] are essentially a result of a supply of money that is too small, and to that extent are monetary phenomena...Complaints about excessive habits of saving are in such circumstances calculated to confuse the mind of the public and to distract attention from the shortcomings of monetary policy."

(Gustav Cassel, Theory of Social Economy, 1918: p.445).

"The present writer is in agreement with the conclusion that the cause of the Great Depression must be sought elsewhere than in savings-investment relationships...[T]he emphasis of contemporary economists on these relationships has been misplaced, because a far more potent factor of economic instability in recent years, namely, erratic variation in the quantity of money, has been ignored."

(Clark Warburton, "The Misplaced Emphasis", Journal of Business, 1946: p.284)

"At last, I have discovered the cause of Christmas!"

(attributed to Nicholas Kaldor, after noting that the money supply surges in December and declines in January)

"Another difference between Milton and myself is that everything reminds Milton of the money supply. Well, everything reminds me of sex, but I keep it out of the paper."

(attributed to Robert M. Solow)



(A) The Money Demand Studies
(B) The Monetary History
(C) The Sims Exercises


The idea that nominal output is largely driven by changes in money supply was the heart of the first controversy between Keynesians and Monetarists. As we saw in our discussion of the Monetarist transmission mechanism, if we take the vertical LM characterization of Monetarism, then only money can drive nominal output; if we take a regular LM characterization, then we can only say that money has a very, very strong effect on nominal output. In either case, the gist of Friedman's (1956) message was clear: money matters. To 1950s Keynesians raised on income-expenditure models, Friedman's assertions on the importance of money reminded Keynesians that they had an LM side which had hitherto been relegated to a faint, supporting role. To this extent, then, Friedman's work was illuminating for the Keynesians, but not necessarily controversial.

The controversy came with Friedman's empirical challenge: namely, to demonstrate that the Monetarist transmission mechanism was empirically dominant. In other words, he sought to show that a lot of the fluctuations in nominal income were largely generated by fluctuations in the supply of money. To defend their case, the Monetarists had to rule out two alternative sources of fluctuations or rather show them to be empirically weak: (1) aggregate demand fluctuations and (2) money demand fluctuations.

(A) The Money Demand Studies

Ruling out money demand fluctuations is particularly crucial. Note that even if the Monetarist transmission mechanism holds true, whether in vertical or regular LM form, this does not necessarily imply that the money supply fluctuations drive output fluctuations: it is equally possible that output fluctuations are generated by money demand fluctuations. Consequently, for the early Monetarists, it became necessary to prove that money demand is empirically stable, by which we mean that it is a stable function of interest, prices, wealth (human and non-human) and perhaps a few other variables. It is important not to be confused about this: a stable money demand does not imply constant velocity!

The stability of the money demand function has been doubted by many Keynesian economists (e.g. Kaldor, 1970; Modigliani, 1977). For the Monetarists, it was of central importance to demonstrate this empirically. This was the heart of the efforts of early Monetarist studies - e.g. Phillip Cagan (1956), Milton Friedman (1959, 1966), Allan H. Meltzer (1963), Karl Brunner and A.H. Meltzer (1963), David Laidler (1966), Gregory Chow (1966), etc.

Despite these early successes, the empirical evidence has since become mixed. In particular, money demand relationships broke down empirically in the 1970s (the classical article being the "case of the missing money" of Goldfeld (1976)). Some sense of stability for money demand relationships has re-emerged in the 1980s for some money aggregates, but it is far from comfortable. As a result, some Monetarists have modified the theory of money demand in order to account for these puzzling empirical findings - most famously, the "buffer stock" theory of money demand set out by Carr and Darby (1981) and Laidler (1984) - but instability of money demand has persisted empirically. For a survey of this research, see Laidler (1977) and Judd and Scadding (1982).

(B) The Monetary History

The next step was to show that money supply fluctuations cause output fluctuations empirically. Gathering evidence to this end had already been the primary research of Clark Warburton (e.g. 1946, 1966). The classic effort was the monumental tome Milton Friedman wrote with Anna J. Schwartz, A Monetary History of the United States (1963). This study was a "historical" rather than "econometric" attempt at empirical analysis. In other words, they examine several particularly pertinent historical episodes of economic upturns and downturns and try to trace their sources to phenomena that preceded them - such as monetary policy decisions, bank panics, etc. - that would have led to substantial changes in the money supply. The approach was rather relaxed and discursive, intermixing the story and the numbers. They made some rather crude estimates of the lag between the money supply fluctuations and the nominal output fluctuations. Naturally, they weaved the Monetarist theory in and out of the data and historical analysis.

Friedman and Schwartz's (1963) methodology was disparaged by contemporary economists. After all, Friedman and Schwartz seem to pursue their empirical claims like "criminal prosecutors", selectively assembling a disparate collection of supporting historical evidence and discarding considerations that do not seem to fit their case, rather than employing formal econometric techniques. In an accompanying article on short-term changes (Friedman and Schwartz, 1963) and two later follow-up treatises on this theme (Friedman and Schwartz, 1970, 1982) they attempt to be a bit more formal. However, the Friedman and Schwartz studies have been subjected to a good degree of hot criticism for their "semiprofessional" methodology, and they were accused of blatant "chicanery" and even "charlatanism" (e.g. Tobin, 1965, 1970; Johnson, 1971; Kaldor, 1970, 1982; Temin, 1976; Hirsch and de Marchi, 1986; Hendry and Ericsson, 1991).

Nonetheless, in tracking down several important episodes in monetary history, Friedman and Schwartz (1963) found that income expansions/contracts were always preceded by expansions/contractions in the money supply. One of their more celebrated hypotheses was that the Great Depression of the 1930s was not the result of insufficient aggregate demand but rather that it resulted from a fall in the supply of money, the result of a misconceived contractionary Federal Reserve monetary policy compounded by the waves of bank failures during the period and a subsequent failure of the Federal Reserve to react appropriately.

There is, of course, nothing outlandish about Friedman-Schwartz's claim that money supply contractions can lead to output contractions. Keynesians could comfortably accept this: after all, a left-ward shift in the LM curve does reduce output, whether the Monetarist transmission mechanism is there or not. Where is the debate?

Ostensibly, as noted, the bone of contention was the relative importance of money supply. The controversial step taken by the Monetarists was to show that money supply fluctuations outperformed the traditional Keynesian drivers of aggregate demand fluctuations in predicting output changes. A comparison between the explanatory power of autonomous investment and money supply was undertaken in a famous study by Milton Friedman and David Meiselman (1963) who found, indeed, that money supply did better (albeit, see Ando and Modigliani (1965) for a response).

This was followed up by the famous "St. Louis equation", a simplistic reduced-form regression which regresses output linearly on current and past values of variables such as the money supply and government spending. The "St Louis" researchers (Anderson and Jordan, 1968; Carlson, 1986) demonstrated that monetary variables outperformed fiscal variables - at least until the 1970s and, consequently, concluded that the Monetarist thesis was superior. However, the form of the St Louis equation engendered much criticism for its oversimplified reduced-form, atheoretical construction and its inability to determine "causality" or differentiate between the Monetarist and Keynesian theses (cf. Kareken and Solow, 1963). Brunner and Meltzer (1976) achieved a somewhat more modest conclusion: namely, they noted that fiscal policy does affect output, but money-financed fiscal policy has a far greater impact than bond-financed fiscal policy - although this is perfectly consistent with Neo-Keynesian theory.

The Friedman and Schwartz (1963) study touched on particularly sore points for the Keynesians: notably, by challenging their favorite examples of aggregate demand failures such as the Great Depression. The Keynesians responded in various fashions. Nicholas Kaldor (1970) took issue with the exogeneity/endogeneity of the money supply. Kaldor proposed that money supply tended to accommodate money demand. This, he notes, would be the case if, for instance, the Federal Reserve desired to stabilize interest rates, or ensure government debt financing, or by the financial intermediaries' own desires and inventive abilities, etc. If money supply followed money demand in this fashion, then velocity would be constant. Changes in aggregate demand - and hence money demand - would feed into changes in money supply. The Friedman-Schwartz "fact" that money supply changes precede output changes would be perfectly compatible with this idea: after all, the Keynesian multiplier (i.e. changes in output) takes time to work itself through. It is entirely plausible that money supply would immediately increase in response to aggregate demand (and thus money demand) but that the resulting output change from the multiplier would only follow a period or two later. Nothing "anti-Keynesian" in that result and perfectly compatible with Friedman and Schwartz's evidence.

Some Monetarists were willing to accept the endogeneity of money supply - or at least that it was not completely autonomous (cf. Brunner, 1968, 1981; Brunner and Meltzer, 1968, 1970). However, the work of Phillip Cagan (1965) was designed to prove the exogeneity of the money supply. One can temper this assertion by arguing that the money supply endogeneity hypothesis of Kaldor et al. is not watertight or, rather, it does not establish a necessary one-way causality between income and money. Its impact can also be ambiguous in the sense that it is possible that a collapse in aggregate demand may lead to a collapse in money supply or an increase in money supply by various endogenous money supply channels. But, the Monetarists claim, the reverse never happens: i.e. money supply contractions never precede output expansions. Consequently, while accepting some degree of endogeneity, the Monetarists could at least argue that the Friedman-Schwartz causality is dominant, or at least more certain. At any rate, as the policy work of some Monetarists (e.g. Friedman, 1959; Brunner and Meltzer, 1983) seems to imply, Monetarists do not really believe the money supply is exogenous inasmuch as they believe that it should be exogenous.

In a celebrated but rather cheeky paper, James Tobin (1970) attempted to show that, nonetheless, the Monetarists had fallen into a "post hoc ergo propter hoc" fallacy, i.e. the fallacy of deriving causation from correlation. Tobin demonstrated that even an "ultra-Keynesian" model (vertical IS, horizontal LM) could generate the kind of timing Friedman and Schwartz found. In this ultra-Keynesian model, money has no effect on income whatsoever, and yet it generates cycles where money supply changes precede output changes. Tobin's trick was to show that as the government budget deficit is reduced in expansion (due to rising tax revenue), government debt is retired very quickly and thus the wealth demand for money - and consequently, by endogeneity, money supply - declines quite fast. As output takes time to catch up to fill the gap, then money supply changes seem to precede output changes. Tobin then went on mischievously to construct an "ultra-Monetarist" model where money demand is related to permanent income and where money supply affects income. Yet, Tobin shows, this "ultra-Monetarist" model nonetheless yields time lags which are, in fact, incompatible with the Friedman-Schwartz evidence. The important lesson Tobin (1970) draws out, then, is simply that "correlation does not imply causation" and that the Monetarists ought to be more careful about what they infer from the Friedman and Schwartz evidence.

At another level, Peter Temin (1976) launched an attack on the Friedman-Schwartz account of the Great Depression. If the Friedman-Schwartzthesis of money contraction were true, Temin asserted, then one ought to have seen interest rates rising. However, he found that there were in fact substantial declines in interest rates during this period. In Temin's view, only an aggregate demand contraction (the Keynesian argument) would be compatible with the falling output and falling interest rates evident in the Great Depression. However, Temin's argument relies on a vertical LM characterization of Monetarism. As noted earlier, a regular LM interpretation would allow for more ambiguous interest rate movements - thus the Monetarist thesis would still be compatible with Temin's evidence. See Schwartz (1981) for a reply and a summary of the resulting debate.

(C) The Sims Exercises

The empirical race took an interesting turn with the famous tests of Christopher A. Sims (1972). Specifically, Sims used the then-novel "Granger-causality" test for a pair of time series introduced by C.W.J. Granger (1969). The method effectively takes one variable, M and regresses it on its own lagged values and the lagged values of another variable Y. Thus, serial correlation in the pair of variables are "washed out" and all that remains is the correlation between them. If the coefficients attached to the lagged values of Y are significant, then Y is said to "Granger-cause" M. To verify the opposite possibility, one regresses Y on its own lagged values and lagged values of the other variable M; if the coefficients attached to M are positive, then M is said to "Granger-cause" Y. Sims (1972) showed using this method that money supply "Granger-caused" income but that income did not "Granger-cause" the money supply - thus lending support to the Monetarist camp.

However, the Granger-causality method only allows one to examine two variables at a time, thus omitting possible common influences from other variables lurking in the background - such as interest rates. In other words, Granger-causality cannot get rid of spurious correlations or correlations driven by a hidden "third factor". A new method, "vector autoregression" (VAR), was introduced during the 1970s which basically applies the logic of Granger-causality to a vector of variables - thus enabling one to handle more than just a pair of variables to examine "causality". Using VAR with interest rates and prices included as well as money and income, Sims (1980) found that while pre-war cycles do seem to support the Monetarist thesis, the post-war cycles are quite different. Specifically, Sims found that in the post-war period, the interest rate accounted for most of the effect on output previously attributed to money. In effect, money became endogenous. The Monetarist hopes were defeated.

backBack topTop Selected References


Home Alphabetical Index Schools of Thought Surveys and Essays
Web Links References Contact Frames